On-line vehicle classification

For various applications it is useful to automatically divide vehicles passing a point into several categories. For the method described two inductive loop detectors were used in each lane. The shape of the bottom of a passing car can be assessed by examining the detector\´s analog signal. This signal, called the "signature," is sampled and fed into a digital computer, which calculates the vehicle\´s length and some shape factors. The passing vehicles can be separated into five categories: 1) passenger car, 2) delivery van, 3) truck, 4) truck-trailer, and 5) truck-semitrailer. To develop and optimize the method, a "learning set" of 1400 signatures was obtained from known vehicles. To investigate the effectiveness of the method, a "test set" of 950 signatures was gathered. By the use of the classifiers which were found with the above procedure, an on-line program in assembler code was written for a PDP-11/20 minicomputer. The determination of the class can be carried out in approximately 500 µs after a vehicle passes the detector. A trial with the on-line system gave very satisfactory results.